0
$\begingroup$

I collected stress data (cortisol reactivity) on 4 timepoints. I also collected psychological measures at 1 timepoint. I would like to know if the psychological measure (one measurement time point) is related to stress reactivity. I tried to calculate correlation coefficients between psychological measure and Cortisol reactivity time 1 psychological measure and Cortisol reactivity time 2 psychological measure and Cortisol reactivity time 3 psychological measure and Cortisol reactivity time 4

Is there another way to deal with these data to know whether the psychological measure has an effect on cortisol reactivity (without having 4 Pearson correlations where I only can see that there are associations at time 1 and 3 but not at time 2 and 4)? Is an repeated measure Ancova (Factor: cortisol reactivity time 1 to time 4 and the psychological measure as covariate?) an alternative? Another option I am thinking about is canonical correlation (Set 1: Cortisol reactivity Time 1, 2, 3,4) And Set 2(one variable of the psychological measure) n=22 individuals.

Thank you very much for your help. It would be great if I get some hints.

Improve this question
$\endgroup$
2
  • $\begingroup$ Your language is a little imprecise. When you say "has an effect on" do you mean correlated to, or that the one thing influences the other in some causal manner. The first can be tested statistically, but the second requires careful study design and lots of controls. $\endgroup$
    – Michael Lew
    Commented Dec 5, 2023 at 20:11
  • $\begingroup$ Sorry i mean correlational association. How would you test correlational associations? $\endgroup$
    – Michi19
    Commented Dec 5, 2023 at 21:12

1 Answer 1

Reset to default
0
$\begingroup$

I hope I understand correctly, but I would personally use a mixed-effect model, since 4 measures of cortisol reactivity are taken on 1 individual and are therefore not independent. For this, you would need to have one column in which you write the individual ID. Each individual has one unique ID. Then one column in which you write the value for the cortisol reactivity. Then one column in which you write at which time point it was (this column will contain values 1, 2, 3 and 4 for each individual. This means that each individual will have 4 lines in the table, each having the same value for individual ID.
In the last column there would be the value for the psychological measure, which will be the same for each line of one individual.

I would personally want to know whether the time point affects the reactivity.

Then I would write a model which in R would look something like : cortisol reactivity~psychological measure*time+(1|individual ID)

What kind of model you chose (linear mixed or generalized linear mixed) would depend on the structure of your data.

Improve this answer
$\endgroup$
13
  • $\begingroup$ Thanks very much for your help. Is it possible to use mixed effect models with 22 individuals and 4 timepoints? I thought that much more data are needed in multilevel models? But it would be great if mixed effect models can be used with my amount of data $\endgroup$
    – Michi19
    Commented Dec 5, 2023 at 17:09
  • $\begingroup$ I forgot to mention that the psychological measure is continuous (working memory indicator) and not ill vs health. So i have 22 x 4timepoints 88 observations. $\endgroup$
    – Michi19
    Commented Dec 6, 2023 at 5:58
  • $\begingroup$ Sorry I mixed something up, I deleted my previous comment. The best thing is you give it a try and you tell us how it went. The thing now is to chose whether it is going to be a linear mixed model or a generalized linear mixed model. May I know why you chose 4 time points and why these time points ? $\endgroup$
    – CaroZ
    Commented Dec 6, 2023 at 11:14
  • $\begingroup$ You could upload a plot of cortisol reactivity according to time to give us an idea :) $\endgroup$
    – CaroZ
    Commented Dec 6, 2023 at 11:31
  • $\begingroup$ Thank you very much. I will try it and i will post a plot. We have 4 timepoint beacsue the participants joined were working on a stressfull task on 4 days (each week). This was necessary because he have some more measures (performance measures etc.) and we wanted to know who performance changes over the 4 timepoints. More timepoint would be preferable but was not possible due to organisational reasons. Thanks very much i will try it and inform you how it works. $\endgroup$
    – Michi19
    Commented Dec 7, 2023 at 7:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.